def test_fit_predict_pipeline(teardown):
x_data = iris.data
y_t_data = iris.target
random_state = 123
n_components = 2
# baikal way
x = Input()
y_t = Input()
x_pca = PCA(n_components=n_components, random_state=random_state, name="pca")(x)
y = LogisticRegression(
multi_class="multinomial",
solver="lbfgs",
random_state=random_state,
name="logreg",
)(x_pca, y_t)
model = Model(x, y, y_t)
y_pred_baikal = model.fit(x_data, y_t_data).predict(x_data)
# traditional way
pca = PCA(n_components=n_components, random_state=random_state)
logreg = LogisticRegression(
multi_class="multinomial", solver="lbfgs", random_state=random_state
)
x_data_transformed = pca.fit_transform(x_data)
y_pred_traditional = logreg.fit(x_data_transformed, y_t_data).predict(
x_data_transformed
)
assert_array_equal(y_pred_baikal, y_pred_traditional)
def test_fit_predict_ensemble(teardown):
mask = iris.target != 2 # Reduce to binary problem to avoid ConvergenceWarning
x_data = iris.data
y_t_data = iris.target
random_state = 123
# baikal way
x = Input()
y_t = Input()
y1 = LogisticRegression(random_state=random_state)(x, y_t)
y2 = RandomForestClassifier(random_state=random_state)(x, y_t)
features = Stack(axis=1)([y1, y2])
y = LogisticRegression(random_state=random_state)(features, y_t)
model = Model(x, y, y_t)
model.fit(x_data, y_t_data)
y_pred_baikal = model.predict(x_data)
# traditional way
logreg = sklearn.linear_model.LogisticRegression(random_state=random_state)
logreg.fit(x_data, y_t_data)
logreg_pred = logreg.predict(x_data)
random_forest = sklearn.ensemble.RandomForestClassifier(random_state=random_state)
random_forest.fit(x_data, y_t_data)
random_forest_pred = random_forest.predict(x_data)
features = np.stack([logreg_pred, random_forest_pred], axis=1)
ensemble = sklearn.linear_model.LogisticRegression(random_state=random_state)
ensemble.fit(features, y_t_data)
y_pred_traditional = ensemble.predict(features)
assert_array_equal(y_pred_baikal, y_pred_traditional)
def test_fit_params(teardown):
x_data = iris.data
y_t_data = iris.target
random_state = 123
n_components = 2
sample_weight = y_t_data + 1 # Just weigh the classes differently
fit_params = {"logreg__sample_weight": sample_weight}
# baikal way
x = Input()
y_t = Input()
x_pca = PCA(n_components=n_components, random_state=random_state, name="pca")(x)
y = LogisticRegression(
multi_class="multinomial",
solver="lbfgs",
random_state=random_state,
name="logreg",
)(x_pca, y_t)
model = Model(x, y, y_t)
model.fit(x_data, y_t_data, **fit_params)
# traditional way
pca = PCA(n_components=n_components, random_state=random_state)
logreg = LogisticRegression(
multi_class="multinomial", solver="lbfgs", random_state=random_state
)
pipe = Pipeline([("pca", pca), ("logreg", logreg)])
pipe.fit(x_data, y_t_data, **fit_params)
# Use assert_allclose instead of all equal due to small numerical differences
# between fit_transform(...) and fit(...).transform(...)
assert_allclose(model.get_step("logreg").coef_, pipe.named_steps["logreg"].coef_)
def test_grid_search_cv_with_tunable_step():
param_grid = {
"classifier": [
LogisticRegression(random_state=random_state),
RandomForestClassifier(random_state=random_state),
],
"pca__n_components": [2, 4],
}
# baikal way
def build_fn():
x = Input()
y_t = Input()
h = PCA(random_state=random_state, name="pca")(x)
y = LogisticRegression(random_state=random_state,
name="classifier")(h, y_t)
model = Model(x, y, y_t)
return model
sk_model = SKLearnWrapper(build_fn)
gscv_baikal = GridSearchCV(
sk_model,
param_grid,
cv=cv,
scoring="accuracy",
return_train_score=True,
verbose=verbose,
)
gscv_baikal.fit(x_data, y_t_data)
# traditional way
pca = PCA(random_state=random_state)
classifier = LogisticRegression(random_state=random_state)
pipe = Pipeline([("pca", pca), ("classifier", classifier)])
gscv_traditional = GridSearchCV(
pipe,
param_grid,
cv=cv,
scoring="accuracy",
return_train_score=True,
verbose=verbose,
)
gscv_traditional.fit(x_data, y_t_data)
assert gscv_baikal.best_params_ == gscv_traditional.best_params_
assert_array_equal(
gscv_traditional.cv_results_["mean_train_score"],
gscv_baikal.cv_results_["mean_train_score"],
)
assert_array_equal(
gscv_traditional.cv_results_["mean_test_score"],
gscv_baikal.cv_results_["mean_test_score"],
)
def test_with_unnecessary_target(self, teardown):
x = Input()
y_t = Input()
logreg = LogisticRegression()
y_p = logreg(x, y_t)
model = Model(x, y_p, y_t)
model.fit(iris.data, iris.target)
# won't require the target is trainable was set to False,
# but won't complain if it was passed to fit
logreg.trainable = False
model.fit(iris.data, iris.target)
def make_naive_stacked_model(n_components, random_state, x_data, y_t_data):
# An unnecessarily complex Model
# Sub-model 1
x1 = Input(name="x1")
y1_t = Input(name="y1_t")
h1 = PCA(n_components=n_components, random_state=random_state, name="pca_sub1")(x1)
y1 = LogisticRegression(
multi_class="multinomial",
solver="lbfgs",
random_state=random_state,
name="logreg_sub1",
)(h1, y1_t)
submodel1 = Model(x1, y1, y1_t, name="submodel1")
# Sub-model 2 (a nested stacked model)
x2 = Input(name="x2")
y2_t = Input(name="y2_t")
y2_1 = RandomForestClassifier(random_state=random_state, name="rforest_sub2")(
x2, y2_t
)
y2_2 = ExtraTreesClassifier(random_state=random_state, name="extrees_sub2")(
x2, y2_t
)
features = Stack(axis=1, name="stack_sub2")([y2_1, y2_2])
y2 = LogisticRegression(
multi_class="multinomial",
solver="lbfgs",
random_state=random_state,
name="logreg_sub2",
)(features, y2_t)
submodel2 = Model(x2, y2, y2_t, name="submodel2")
# Stack of submodels
x = Input(name="x")
y_t = Input(name="y_t")
y1 = submodel1(x, y_t)
y2 = submodel2(x, y_t)
features = Stack(axis=1, name="stack")([y1, y2])
y = LogisticRegression(
multi_class="multinomial",
solver="lbfgs",
random_state=random_state,
name="logreg_stacked",
)(features, y_t)
stacked_model_baikal = Model(x, y, y_t, name="stacked")
stacked_model_baikal.fit(x_data, y_t_data)
return stacked_model_baikal
def test_fit_predict_standard_stack(teardown):
# This uses the "standard" protocol where the 2nd level features
# are the out-of-fold predictions of the 1st. It also appends the
# original data to the 2nd level features.
# See for example: https://www.kdnuggets.com/2017/02/stacking-models-imropved-predictions.html
X_data, y_t_data = breast_cancer.data, breast_cancer.target
X_train, X_test, y_t_train, y_t_test = train_test_split(X_data,
y_t_data,
test_size=0.2,
random_state=0)
random_state = 42
# baikal way
x = Input()
y_t = Input()
y_p1 = RandomForestClassifierOOF(n_estimators=10,
random_state=random_state)(
x, y_t, compute_func="predict_proba")
y_p1 = Lambda(lambda array: array[:, 1:])(y_p1) # remove collinear feature
x_scaled = StandardScaler()(x)
y_p2 = LinearSVCOOF(random_state=random_state)(
x_scaled, y_t, compute_func="decision_function")
stacked_features = ColumnStack()([x, y_p1, y_p2])
y_p = LogisticRegression(solver="liblinear",
random_state=random_state)(stacked_features, y_t)
model = Model(x, y_p, y_t)
model.fit(X_train, y_t_train)
y_pred_baikal = model.predict(X_test)
# traditional way
estimators = [
("rf",
RandomForestClassifier(n_estimators=10, random_state=random_state)),
("svr",
make_pipeline(StandardScaler(),
LinearSVC(random_state=random_state))),
]
clf = sklearn.ensemble.StackingClassifier(
estimators=estimators,
final_estimator=LogisticRegression(solver="liblinear",
random_state=random_state),
passthrough=True,
)
y_pred_traditional = clf.fit(X_train, y_t_train).predict(X_test)
assert_array_equal(y_pred_baikal, y_pred_traditional)
def test_with_steps_with_duplicated_names(self, teardown):
x = Input()
h = PCA(name="duplicated-name")(x)
y = LogisticRegression(name="duplicated-name")(h)
with pytest.raises(RuntimeError):
Model(x, y)
def dataplaceholders(self):
x1 = Input(name="x1")
x2 = Input(name="x2")
y1_t = Input(name="y1_t")
y1 = LogisticRegression()(x1, y1_t)
y2 = PCA()(x2)
return x1, x2, y1, y2, y1_t
def test_nested_model(teardown):
x_data = iris.data
y_t_data = iris.target
# Sub-model
x = Input()
y_t = Input()
h = PCA(n_components=2)(x)
y = LogisticRegression()(h, y_t)
submodel = Model(x, y, y_t)
# Model
x = Input()
y_t = Input()
y = submodel(x, y_t)
model = Model(x, y, y_t)
with raises_with_cause(RuntimeError, NotFittedError):
submodel.predict(x_data)
model.fit(x_data, y_t_data)
y_pred = model.predict(x_data)
y_pred_sub = submodel.predict(x_data)
assert_array_equal(y_pred, y_pred_sub)
def test_plot_independent_submodels(teardown, tmp_path, expand_nested):
xs, y_ts, y_ps, steps = [], [], [], []
for i in range(3):
step = (LogisticRegression() if i == 0 else build_submodel(
LogisticRegression(), i))
x = Input(name="x{}".format(i))
y_t = Input(name="y_t{}".format(i))
y_p = step(x, y_t)
xs.append(x)
y_ts.append(y_t)
y_ps.append(y_p)
model = Model(xs, y_ps, y_ts)
filename = str(tmp_path / "test_plot_model.png")
plot_model(model, filename, show=False, expand_nested=expand_nested)
def build_fn():
x = Input()
y_t = Input()
h = PCA(random_state=random_state, name="pca")(x)
y = LogisticRegression(random_state=random_state,
name="classifier")(h, y_t)
model = Model(x, y, y_t)
return model
def test_with_undefined_target(self, teardown):
x = Input()
y = LogisticRegression()(x, trainable=True)
model = Model(inputs=x, outputs=y)
with raises_with_cause(RuntimeError, TypeError):
# LogisticRegression.fit will be called with not enough arguments
# hence the TypeError
model.fit(iris.data)
def dataplaceholders(self):
x1 = Input(name="x1")
x2 = Input(name="x2")
y1_t = Input(name="y1_t")
x1_rescaled = StandardScaler()(x1)
y1 = LogisticRegression()(x1_rescaled, y1_t)
y2 = PCA()(x2)
return x1, x2, x1_rescaled, y1, y2, y1_t
def test_set_params(self, teardown):
step = LogisticRegression()
new_params_wrong = {"non_existent_param": 42}
with pytest.raises(ValueError):
step.set_params(**new_params_wrong)
new_params = {"C": 100.0, "fit_intercept": False, "penalty": "l1"}
step.set_params(**new_params)
params = step.get_params()
expected = {
"C": 100.0,
"class_weight": None,
"dual": False,
"fit_intercept": False,
"intercept_scaling": 1,
"max_iter": 100,
"multi_class": "warn",
"n_jobs": None,
"penalty": "l1",
"random_state": None,
"solver": "warn",
"tol": 0.0001,
"verbose": 0,
"warm_start": False,
"l1_ratio": None,
}
assert expected == params
def test_compute_func(self, simple_step, shared_step, dataplaceholders, teardown):
assert simple_step.compute_func == simple_step.predict
simple_step.compute_func = simple_step.predict_proba
assert simple_step.compute_func == simple_step.predict_proba
with pytest.raises(AttributeError):
shared_step.compute_func
with pytest.raises(AttributeError):
shared_step.compute_func = shared_step.predict_proba
with pytest.raises(AttributeError):
# because the step hasn't been called
LogisticRegression().compute_func
with pytest.raises(AttributeError):
# because the step hasn't been called
LogisticRegression().compute_func = lambda x: x
def test_trainable(self, simple_step, shared_step, dataplaceholders, teardown):
assert simple_step.trainable
simple_step.trainable = False
assert not simple_step.trainable
with pytest.raises(AttributeError):
shared_step.trainable
with pytest.raises(AttributeError):
shared_step.trainable = True
with pytest.raises(AttributeError):
# because the step hasn't been called
LogisticRegression().trainable
with pytest.raises(AttributeError):
# because the step hasn't been called
LogisticRegression().trainable = False
def build_fn():
x = Input()
y_t = Input()
h = PCA(random_state=random_state, name="pca")(x)
y = LogisticRegression(random_state=random_state,
solver="liblinear",
name="logreg")(h, y_t)
model = Model(x, y, y_t)
return model
def test_inputs(self, simple_step, shared_step, dataplaceholders, teardown):
x1 = dataplaceholders[0]
assert simple_step.inputs == [x1]
with pytest.raises(AttributeError):
shared_step.inputs
with pytest.raises(AttributeError):
# because the step hasn't been called
LogisticRegression().inputs
def test_predict_with_not_fitted_steps(self, teardown):
x_data = iris.data
x = Input(name="x")
xt = PCA(n_components=2)(x)
y = LogisticRegression(multi_class="multinomial", solver="lbfgs")(xt)
model = Model(x, y)
with raises_with_cause(RuntimeError, NotFittedError):
model.predict(x_data)
def test_outputs(self, simple_step, shared_step, dataplaceholders, teardown):
*_, y_simple, y_shared_1, y_shared_2 = dataplaceholders
assert simple_step.outputs == [y_simple]
with pytest.raises(AttributeError):
shared_step.outputs
with pytest.raises(AttributeError):
# because the step hasn't been called
LogisticRegression().outputs
def test_targets(self, simple_step, shared_step, dataplaceholders, teardown):
y_t = dataplaceholders[2]
assert simple_step.targets == [y_t]
with pytest.raises(AttributeError):
shared_step.targets
with pytest.raises(AttributeError):
# because the step hasn't been called
LogisticRegression().targets
def test_grid_search_cv():
param_grid = {
"pca__n_components": [2, 4],
"logreg__C": [0.1, 1.0, 10],
"logreg__penalty": ["l1", "l2"],
}
# baikal way
def build_fn():
x = Input()
y_t = Input()
h = PCA(random_state=random_state, name="pca")(x)
y = LogisticRegression(random_state=random_state,
solver="liblinear",
name="logreg")(h, y_t)
model = Model(x, y, y_t)
return model
sk_model = SKLearnWrapper(build_fn)
assert isinstance(sk_model.model, Model)
gscv_baikal = GridSearchCV(
sk_model,
param_grid,
cv=cv,
scoring="accuracy",
return_train_score=True,
verbose=verbose,
)
gscv_baikal.fit(x_data, y_t_data)
# traditional way
pca = PCA(random_state=random_state)
logreg = LogisticRegression(random_state=random_state, solver="liblinear")
pipe = Pipeline([("pca", pca), ("logreg", logreg)])
gscv_traditional = GridSearchCV(
pipe,
param_grid,
cv=cv,
scoring="accuracy",
return_train_score=True,
verbose=verbose,
)
gscv_traditional.fit(x_data, y_t_data)
assert gscv_baikal.best_params_ == gscv_traditional.best_params_
assert_array_equal(
gscv_traditional.cv_results_["mean_train_score"],
gscv_baikal.cv_results_["mean_train_score"],
)
assert_array_equal(
gscv_traditional.cv_results_["mean_test_score"],
gscv_baikal.cv_results_["mean_test_score"],
)
def test_with_non_fitted_non_trainable_step(self, teardown):
x = Input()
y_t = Input()
z = PCA()(x, trainable=False)
y = LogisticRegression()(z, y_t)
model = Model(x, y, y_t)
with raises_with_cause(RuntimeError, NotFittedError):
# this will raise an error when calling compute
# on PCA which was flagged as trainable=False but
# hasn't been fitted
model.fit(iris.data, iris.target)
def test_with_unnecessary_inputs(self, teardown):
x1 = Input()
x2 = Input()
y_t = Input()
h = PCA()(x1)
y = LogisticRegression()(h, y_t)
with pytest.raises(ValueError):
Model([x1, x2], y, y_t)
with pytest.raises(ValueError):
Model([x1, h], y, y_t) # x1 is an unnecessary input upstream of h
def test_fit_predict_ensemble_with_proba_features(teardown):
mask = iris.target != 2 # Reduce to binary problem to avoid ConvergenceWarning
x_data = iris.data[mask]
y_t_data = iris.target[mask]
random_state = 123
n_estimators = 5
# baikal way
x = Input()
y_t = Input()
y1 = LogisticRegression(random_state=random_state, function="predict_proba")(x, y_t)
y2 = RandomForestClassifier(
n_estimators=n_estimators, random_state=random_state, function="apply"
)(x, y_t)
features = Concatenate(axis=1)([y1, y2])
y = LogisticRegression(random_state=random_state)(features, y_t)
model = Model(x, y, y_t)
model.fit(x_data, y_t_data)
y_pred_baikal = model.predict(x_data)
# traditional way
logreg = sklearn.linear_model.LogisticRegression(random_state=random_state)
logreg.fit(x_data, y_t_data)
logreg_proba = logreg.predict_proba(x_data)
random_forest = sklearn.ensemble.RandomForestClassifier(
n_estimators=n_estimators, random_state=random_state
)
random_forest.fit(x_data, y_t_data)
random_forest_leafidx = random_forest.apply(x_data)
features = np.concatenate([logreg_proba, random_forest_leafidx], axis=1)
ensemble = sklearn.linear_model.LogisticRegression(random_state=random_state)
ensemble.fit(features, y_t_data)
y_pred_traditional = ensemble.predict(features)
assert_array_equal(y_pred_baikal, y_pred_traditional)
def test_get_params(self, teardown):
step = LogisticRegression()
params = step.get_params()
expected = {
"C": 1.0,
"class_weight": None,
"dual": False,
"fit_intercept": True,
"intercept_scaling": 1,
"max_iter": 100,
"multi_class": "warn",
"n_jobs": None,
"penalty": "l2",
"random_state": None,
"solver": "warn",
"tol": 0.0001,
"verbose": 0,
"warm_start": False,
"l1_ratio": None,
}
assert expected == params
def test_plot_nested_submodels(teardown, tmp_path, levels, expand_nested):
submodels = [LogisticRegression()]
for level in range(levels):
sub_model = build_submodel(submodels[level], level + 1)
submodels.append(sub_model)
x = Input(name="x")
y_t = Input(name="y_t")
y_p = submodels[-1](x, y_t)
model = Model(x, y_p, y_t)
filename = str(tmp_path / "test_plot_model.png")
plot_model(model, filename, show=False, expand_nested=expand_nested)
def test_with_wrong_type(self, teardown):
x = Input()
y_t = Input()
y = LogisticRegression()(x, y_t)
wrong = np.zeros((10,))
with pytest.raises(ValueError):
Model(wrong, y, y_t)
with pytest.raises(ValueError):
Model(x, wrong, y_t)
with pytest.raises(ValueError):
Model(x, y, wrong)
def test_plot_shared_submodel(teardown, tmp_path, expand_nested):
xs, y_ts, y_ps = [], [], []
submodel = build_submodel(LogisticRegression(), 0)
for i in range(2):
x = Input(name="x{}".format(i))
y_t = Input(name="y_t{}".format(i))
y_p = submodel(x, y_t)
xs.append(x)
y_ts.append(y_t)
y_ps.append(y_p)
model = Model(xs, y_ps, y_ts)
filename = str(tmp_path / PNG_FILENAME)
plot_model(model, filename, show=False, expand_nested=expand_nested)
